Interactive Graph Navigation on IGN Qwen3.5-8B backbone (evaluation)
39SR@1.5 (Easy)TANDEM
Evaluation Results
| Method | Links | |||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| TANDEMBackbone=Qwen3.5-8B, Protocol=Cost-Aware QA, spmem=true, QA=true2026.06 | 39 | 2.32 | 35.8 | 3.13 | 31 | 5.33 | 14.2 | 22.3 | 35.3 | 50 | 66.5 | 72.5 | 2.68 | 23.87 | 0.22 | 0.97 | 1.28 | 1.1 | 3.57 | 21.4 | 40.8 | |
| TANDEM w/o spmemBackbone=Qwen3.5-8B, Protocol=Cost-Aware QA, spmem=false, QA=true2026.06 | 31.2 | 2.1 | 29 | 2.49 | 24 | 4.56 | 8.8 | 16 | 28 | 43.5 | 57.8 | 65 | 3.15 | 24.08 | 0.26 | 1 | 0.82 | 0.93 | 3.01 | 18.5 | 36 | |
| GTABackbone=Qwen3.5-8B, Protocol=Naive QA (unconstrained)2026.06 | 29.7 | 6.93 | 27.6 | 8.17 | 22.4 | 9.76 | 8.8 | 17.6 | 26.7 | 39.3 | 50.6 | 57.9 | 3.49 | 27.62 | 1.73 | 1.46 | 2.87 | 2.19 | 8.25 | 6.4 | 12.2 | |
| GTABackbone=Qwen3.5-8B, Protocol=Cost-Aware QA2026.06 | 29.2 | 2.24 | 27 | 3.01 | 21.8 | 5.14 | 10.5 | 17.2 | 26 | 38.6 | 49.3 | 56.1 | 3.55 | 26.85 | 0.24 | 0.95 | 1.2 | 1.05 | 3.44 | 15.9 | 30.3 | |
| TANDEM w/o QABackbone=Qwen3.5-8B, Protocol=Cost-Aware QA, spmem=true, QA=false2026.06 | 24.5 | — | 17 | — | 17.8 | — | 3.2 | 10.5 | 20 | 34 | 49.2 | 58.4 | 3.35 | 17.27 | — | — | — | — | — | 14.2 | 34.2 | |
| COINBackbone=Qwen3.5-8B, Protocol=Naive QA (unconstrained)2026.06 | 23.6 | 7.48 | 18.7 | 9.03 | 14.3 | 10.81 | 5.9 | 12.4 | 18.6 | 29.2 | 36.8 | 46.3 | 4.63 | 25.18 | 2.04 | 1.17 | 3.58 | 2.27 | 9.06 | 3.7 | 7.3 | |
| MapGPTBackbone=Qwen3.5-8B, Protocol=Naive QA (unconstrained)2026.06 | 20.1 | 7.27 | 17.1 | 8.91 | 12.7 | 10.46 | 6.7 | 12.1 | 15.9 | 28.3 | 32.6 | 41.7 | 4.89 | 29.07 | 1.84 | 1.39 | 3.11 | 2.53 | 8.87 | 3.6 | 7.1 | |
| MapGPTBackbone=Qwen3.5-8B, Protocol=Cost-Aware QA2026.06 | 19.6 | 2.15 | 17.4 | 2.87 | 12.2 | 4.87 | 6.6 | 12.3 | 16.5 | 27.6 | 31.5 | 40.2 | 4.97 | 28.68 | 0.26 | 0.93 | 1.1 | 0.99 | 3.28 | 10.2 | 19.4 | |
| NavGPTBackbone=Qwen3.5-8B, Protocol=Naive QA (unconstrained)2026.06 | 18.1 | 7.34 | 15.6 | 8.99 | 10.9 | 10.68 | 5.1 | 11.4 | 14.8 | 24.6 | 28.9 | 36.7 | 5.23 | 31.64 | 1.99 | 1.18 | 3.24 | 2.61 | 9.02 | 3.1 | 6.1 | |
| COINBackbone=Qwen3.5-8B, Protocol=Cost-Aware QA2026.06 | 18 | 1.37 | 12.5 | 2.19 | 8.5 | 4.33 | 2 | 3.5 | 13 | 22 | 27.4 | 35.9 | 5.2 | 28.4 | 0.45 | 1 | 0 | 1.16 | 2.61 | 9.2 | 19.3 | |
| TANDEM w/o spmem & QABackbone=Qwen3.5-8B, Protocol=Cost-Aware QA, spmem=false, QA=false2026.06 | 18 | — | 14 | — | 10 | — | 1.5 | 6 | 14 | 25 | 34.2 | 44.1 | 5.15 | 21.16 | — | — | — | — | — | 10 | 23.8 | |
| NavGPTBackbone=Qwen3.5-8B, Protocol=Cost-Aware QA2026.06 | 17.5 | 2.11 | 15.3 | 2.82 | 10.1 | 4.76 | 5.8 | 10.8 | 14.3 | 24.1 | 27.8 | 35.3 | 5.31 | 29.15 | 0.27 | 0.92 | 1.06 | 0.96 | 3.21 | 8.9 | 16.9 |